image: Step into the future of pharmaceutical innovation with Insilico
Credit: Insilico Medicine
On December 10, Insilico Medicine, a clinical stage generative artificial intelligence (AI)-driven biotechnology company, hosted the fourth edition of its Pharma.AI Quarterly Launch webinar, titled “Epic Year-End Recap & Q4 Winter Updates”. The event drew more than 300 registrants from universities, healthcare institutions, global pharmaceutical companies, and innovative biotech firms worldwide.
Insilico's software team showcased the recap of Pharm.AI in 2025,and the latest capabilities through live demos and real‑world case studies.
Key highlights are summarized below:
Generative Biologics
What improved in 2025:
- Peptide workflows: template-based screening of peptide libraries against a receptor and manual hotspot-based pocket selection with guided peptide length and 3D analysis.
- Model training: 3D-augmented models with improved performance (Pearson/Spearman +0.2–0.3), validated on the Ginkgo antibody dataset.
- Generative engine: a new diffusion-based antibody model (epitope- and framework-conditioned) and configurable liability constraints with motif-based filtering.
- PDB integration: fetch, auto-fix, and use PDB structures directly for experiments.
What's new in Q4:
- Updated contact visualization and evaluation: Clearer, residue-level contact maps for faster interpretation of target–binder interactions.
- Improved binder scoring: Refined scoring function that better integrates structural and energetic features for more accurate rankings.
- New MD simulation experiment: Molecular dynamics simulations to evaluate binder stability and conformational behavior under realistic biophysical conditions.
Chemistry42/Nach01
What improved in 2025
- Generative chemistry: Protein-based pharmacophore points now complement ligand-based ones, enabling diverse, protein-informed pharmacophore-guided generation.
- Alchemistry 2.0: A two-stage RBFE engine (equilibrium MD + non-equilibrium switching) that delivers industry-level accuracy and massive scalability, with an interactive trajectory viewer and an upcoming absolute binding free energy module. Compared to the previous version, it improves accuracy by up to 16% and achieves a 2–4× speed-up.
- MDFlow: it adds configurable MD simulations, rich trajectory analytics, and multi-format export. Together with Alchemistry 2.0, it forms a continuous physics-based workflow from free energy calculations to full MD.
- ADMET Profiling: it was fully rebuilt with a new training framework, improving key regression endpoints and off-target safety assessment, including 70+ new GPCR inhibitor models, thereby extending Chemistry42’s predictive power into more complex pharmacological spaces.
- Retrosynthesis: The system supports generated, custom, and sketched molecules, uses expert reaction templates and an AI Root Planner to generate selective, literature-consistent synthetic routes, integrates with eMolecules and user building-block libraries for cost-aware planning, and delivers fast, batch retrosynthesis with detailed downloadable reports and CDX scheme export.
- Natch 01: A multimodal foundation model for chemistry and drug discovery. Trained on billions of molecular and textual data points, it combines language understanding and chemical intelligence for property prediction, molecular design, and scientific reasoning. Now available on AWS Marketplace for academic and industrial integration, Natch 01 will soon gain generative capabilities, further embedding it as the core of Chemistry42’s AI-driven ecosystem.
What's new in Q4:
- MolSpace uses Generative Topographic Mapping (GTM) to visualize generated molecules against large public datasets through a streamlined SDF/CSV-to-analysis workflow. New density, classification, regression, and comparison landscape modes, combined with smart sampling and MDS, provide an accurate, interpretable view of both common and rare regions of chemical space.
- PACE (Patent Analysis Chemical Engine) automatically extracts structured chemical data and SAR information from PDF patents by segmenting and recognizing chemical structures and SAR tables, leveraging specialized vision models and visual language models for deeper structure–activity and patent space analysis.
PandaOmics
What improved in 2025:
- Expanded functionality and analytical rigor for target and biomarker discovery.
- Prioritize targets with four new LLM-derived scoring metrics plus an updated ranking framework for more balanced, clinically relevant selection
- The Confidence Score measures the evidence supporting the target's role in disease mechanisms.
- The commercial tractability score assesses the target's market feasibility, taking into account demand and competitive landscape.
- The drug ability score indicates the likelihood of successfully developing the drug.
- The mechanism clarity score examines the understanding of how the intervention affects pathways and diseases.
- 27 scores to analyze data from multi -omics, text and other sources to support target identification and biomarker discovery projects. To enable a smooth start to any analysis. We introduced a new ranking scheme that averages five core scores to highlight well -studied targets with balanced evidence.
What's new in Q4:
- Multi-entry gene support
- Upload and analyze multiple molecular entities (isoforms, transcripts, or protein variants) per a single gene
- Extended species support: Transcriptomic datasets from dog and cat studies can be uploaded, with automatic species recognition when using Ensembl or NCBI Entrez ID.
- Strengthened data security and governance via:
- Virtual Private Cloud (VPC) deployment,
- Single Sign-On (SSO) integration,
- Enhanced data management features,
- Secure intra-organizational sharing of proprietary datasets.
The single-cell dataset viewer and gene signature will come soon.
DORA:
Deep Research compresses the journey from idea to publication‑ready report. After you enter a topic and brief context in a chat interface, it builds a focused research plan and retrieves only high‑quality sources, scored by peer‑review status, author expertise, and relevance, each with a traceable link to the original.
It then applies advanced reasoning to synthesize cross‑domain knowledge, resolve contradictions, and surface hidden questions, exposing a transparent chain of thought behind every conclusion. It produces a concise, analyst‑level report with clearly structured key insights, supporting evidence, and actionable recommendations within minutes. The report opens in DORA, where you can refine, polish, and extend it with built‑in AI tools before using it for decisions or publication.
MMAI GYM for Science:
MMAI GYM is a place where you can send any AI model to train and become an expert in real-world drug development tasks with high-performance. It combines proprietary multi‑omics datasets, patents, papers, clinical readouts, and domain‑specific reinforcement learning to teach models how to work natively with biochemical formats, understand complex R&D tasks and operate constraints of biology, safety, efficacy, and synthesis. Dramatically boosts the biological and chemical intelligence of any causal or frontier LLM, delivering up to 10× performance gains on key drug discovery benchmarks.
In 2016, Insilico first described the concept of using generative AI for designing novel molecules in a peer-reviewed journal, laying the foundation for the commercially available Pharma.AI platform. Since then, Insilico has continued to integrate technical breakthroughs into the Pharma.AI platform, which is now a generative AI-powered solution encompassing biology, chemistry, medicine development, and scientific research. Powered by Pharma.AI, Insilico has nominated over 20 developmental/preclinical candidates (DC/PCC) in its comprehensive portfolio of more than 30 assets since 2021, received IND clearance for 10 molecules, and completed multiple human clinical trials for two of the most advanced pipelines, with positive results announced.
By integrating the technologies of AI and automation, Insilico has demonstrated a significant efficiency boost compared to traditional drug discovery methods (often requiring 2.5-4 years), as announced in the recent key timeline benchmarks for internal DC programs from 2021 to 2024: the average time to DC is 12-18 months, with 60-200 molecules synthesized and tested per program
About Insilico Medicine
Insilico Medicine, a leading and global AI-driven biotech company, utilizes its proprietary Pharma.AI platform and cutting-edge automated laboratory to accelerate drug discovery and advance innovations in life sciences research. By integrating AI and automation technologies and deep in-house drug discovery capabilities, Insilico is delivering innovative drug solutions for unmet needs, including fibrosis, oncology, immunology, pain, obesity, and metabolic disorders. Additionally, Insilico extends the reach of Pharma.AI across various industries, including advanced materials, agriculture, nutritional products, and veterinary medicine.
For more information, please visit www.insilico.com.